这是一个unet-pytorch的源码，可以训练自己的模型

Last update: Jan 05, 2023

Related tags

Overview

Unet：U-Net: Convolutional Networks for Biomedical Image Segmentation目标检测模型在Pytorch当中的实现

性能情况

unet并不适合VOC此类数据集，其更适合特征少，需要浅层特征的医药数据集之类的。

训练数据集	权值文件名称	测试数据集	输入图片大小	mIOU
VOC12+SBD	unet_voc.pth	VOC-Val12	512x512	55.11

所需环境

torch==1.2.0
torchvision==0.4.0

注意事项

unet_voc.pth是基于VOC拓展数据集训练的。
unet_medical.pth是使用示例的细胞分割数据集训练的。
在使用时需要注意区分。

文件下载

训练所需的unet_voc.pth和unet_medical.pth可在百度网盘中下载。
链接: https://pan.baidu.com/s/1AUBpqsSgamoQGEYpNjJg7A 提取码: i3ck

VOC拓展数据集的百度网盘如下：
链接: https://pan.baidu.com/s/1BrR7AUM1XJvPWjKMIy2uEw 提取码: vszf

预测步骤

一、使用预训练权重

a、VOC预训练权重

下载完库后解压，如果想要利用voc训练好的权重进行预测，在百度网盘或者release下载unet_voc.pth，放入model_data，运行即可预测。

img/street.jpg

利用video.py可进行摄像头检测。

b、医药预训练权重

下载完库后解压，如果想要利用医药数据集训练好的权重进行预测，在百度网盘或者release下载unet_medical.pth，放入model_data，修改unet.py中的model_path和num_classes；

_defaults = {
    "model_path"        : 'model_data/unet_voc.pth',
    "model_image_size"  : (512, 512, 3),
    "num_classes"       : 21,
    "cuda"              : True,
    #--------------------------------#
    #   blend参数用于控制是否
    #   让识别结果和原图混合
    #--------------------------------#
    "blend"             : True
}

运行即可预测。

img/cell.png

二、使用自己训练的权重

按照训练步骤训练。
在unet.py文件里面，在如下部分修改model_path、backbone和num_classes使其对应训练好的文件；model_path对应logs文件夹下面的权值文件。

_defaults = {
    "model_path"        : 'model_data/unet_voc.pth',
    "model_image_size"  : (512, 512, 3),
    "num_classes"       : 21,
    "cuda"              : True,
    #--------------------------------#
    #   blend参数用于控制是否
    #   让识别结果和原图混合
    #--------------------------------#
    "blend"             : True
}

运行predict.py，输入

img/street.jpg

利用video.py可进行摄像头检测。

训练步骤

一、训练voc数据集

将我提供的voc数据集放入VOCdevkit中（无需运行voc2unet.py）。
在train.py中设置对应参数，默认参数已经对应voc数据集所需要的参数了，所以只要修改backbone和model_path即可。
运行train.py进行训练。

二、训练自己的数据集

本文使用VOC格式进行训练。
训练前将标签文件放在VOCdevkit文件夹下的VOC2007文件夹下的SegmentationClass中。
训练前将图片文件放在VOCdevkit文件夹下的VOC2007文件夹下的JPEGImages中。
在训练前利用voc2unet.py文件生成对应的txt。
注意修改train.py的num_classes为分类个数+1。
运行train.py即可开始训练。

三、训练医药数据集

下载VGG的预训练权重到model_data下面。
按照默认参数运行train_medical.py即可开始训练。

miou计算

参考miou计算视频和博客。

Reference

https://github.com/ggyyzm/pytorch_segmentation
https://github.com/bonlime/keras-deeplab-v3-plus

Comments

询问一下预训练的问题

你好，打扰了。我是想问下主干模型是指的是在下采样过程中使用的vgg吗？如果我不改变上采样是不是就不用使用imagenet训练。然后注销掉model_path=‘’ 以及 if model_path !=‘’这段。然后使用自己的数据集去进行训练。谢谢大佬！！！！！！。实际上大佬你的voc的权重文件是不是为二次预训练的数据。不好意思，语言表达能力不行。俺不晓得这样说大佬明不明白。

opened by Nine9844 5
训练一段时间后，CE loss变为NAN

您好，看了您的教程我试着自己搭建了一个U-Net模型，并采用Dice + CE loss作为损失函数，但在迭代几十个epoch后，我的CE loss返回了NAN值，反馈的结果是 ‘Function 'LogSoftmaxBackward' returned nan values in its 0th output.’ 同样的数据在您源码上运行没有出现这个问题，请问您是否知道些解决方法？

opened by Breeze-Zero 2
为啥在dataloader第40行转换的array的shape和cv2不一样呢

我使用json_to_dataset.py转化mask后尝试使用代码查看shape import cv2 import numpy as np from PIL import Image

file = '/home/fut/Downloads/unet-pytorch-main/mydata/masks/ID_1110_json.png' img = cv2.imread(file, cv2.IMREAD_UNCHANGED) print(img.shape)

pil = Image.open(file) img2 = np.array(pil) print(img2.shape) 结果会是： (800, 800, 3) (800, 800) 为什么PIL读取后通道就没了，正是因为这个原因你的项目会很好跑起来。

opened by futureflsl 1
from tqdm import tqdm 报错

import os import time

import numpy as np import torch import torch.backends.cudnn as cudnn import torch.optim as optim from torch.utils.data import DataLoader from tqdm import tqdm

opened by Luke-Wei 1

Releases(v3.0)

v3.0(Apr 22, 2022)
重要更新

支持step、cos学习率下降法。

支持adam、sgd优化器选择。

支持不同预测模式的选择，单张图片预测、文件夹预测、视频预测、图片裁剪。

更新summary.py文件，用于观看网络结构。

增加了多GPU训练。

Source code(tar.gz)
Source code(zip)
v2.2(Mar 4, 2022)
重要更新

更新train.py文件，增加了大量的注释，增加多个可调整参数。

更新predict.py文件，增加了大量的注释，增加fps、视频预测、批量预测等功能。

更新unet.py文件，增加了大量的注释，增加先验框选择、置信度、非极大抑制等参数。

合并get_dr_txt.py、get_gt_txt.py和get_map.py文件，通过一个文件来实现数据集的评估。

更新voc_annotation.py文件，增加多个可调整参数。

更新callback.py文件，防止多线程错误。

更新summary.py文件，用于观看网络结构。

Source code(tar.gz)
Source code(zip)
v1.0(Mar 12, 2021)

Source code(tar.gz)
Source code(zip)
unet_resnet_medical.pth(167.85 MB)
unet_resnet_voc.pth(167.85 MB)
unet_vgg_medical.pth(94.96 MB)
unet_vgg_voc.pth(94.96 MB)

Owner

Bubbliiiing

GitHub Repository

This is the official implementation of Elaborative Rehearsal for Zero-shot Action Recognition (ICCV2021)

Elaborative Rehearsal for Zero-shot Action Recognition This is an official implementation of: Shizhe Chen and Dong Huang, Elaborative Rehearsal for Ze

26 Sep 24, 2022

Retinal Vessel Segmentation with Pixel-wise Adaptive Filters (ISBI 2022)

Retinal Vessel Segmentation with Pixel-wise Adaptive Filters (ISBI 2022) Introdu

14 Oct 27, 2022

HEAM: High-Efficiency Approximate Multiplier Optimization for Deep Neural Networks

Approximate Multiplier by HEAM What's HEAM? HEAM is a general optimization method to generate high-efficiency approximate multipliers for specific app

4 Sep 11, 2022

A PyTorch implementation of the continual learning experiments with deep neural networks

Brain-Inspired Replay A PyTorch implementation of the continual learning experiments with deep neural networks described in the following paper: Brain

182 Dec 27, 2022

A lightweight library to compare different PyTorch implementations of the same network architecture.

TorchBug is a lightweight library designed to compare two PyTorch implementations of the same network architecture. It allows you to count, and compar

5 Jan 02, 2023

PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).

PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR)

3k Dec 31, 2022

This project is a loose implementation of paper "Algorithmic Financial Trading with Deep Convolutional Neural Networks: Time Series to Image Conversion Approach"

Stock Market Buy/Sell/Hold prediction Using convolutional Neural Network This repo is an attempt to implement the research paper titled "Algorithmic F

136 Dec 28, 2022

Caffe implementation for Hu et al. Segmentation for Natural Language Expressions

Segmentation from Natural Language Expressions This repository contains the Caffe reimplementation of the following paper: R. Hu, M. Rohrbach, T. Darr

10 Jul 27, 2021

[NeurIPS 2021] Low-Rank Subspaces in GANs

Low-Rank Subspaces in GANs Figure: Image editing results using LowRankGAN on StyleGAN2 (first three columns) and BigGAN (last column). Low-Rank Subspa

112 Dec 28, 2022

Official implementation of VaxNeRF (Voxel-Accelearated NeRF).

VaxNeRF Paper | Google Colab This is the official implementation of VaxNeRF (Voxel-Accelearated NeRF). VaxNeRF provides very fast training and slightl

132 Nov 21, 2022

An Implicit Function Theorem (IFT) optimizer for bi-level optimizations

iftopt An Implicit Function Theorem (IFT) optimizer for bi-level optimizations. Requirements Python 3.7+ PyTorch 1.x Installation $ pip install git+ht

2 Dec 02, 2021

Code for the paper BERT might be Overkill: A Tiny but Effective Biomedical Entity Linker based on Residual Convolutional Neural Networks

Biomedical Entity Linking This repo provides the code for the paper BERT might be Overkill: A Tiny but Effective Biomedical Entity Linker based on Res

24 Oct 24, 2022

DSTC10 Track 2 - Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations

DSTC10 Track 2 - Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations This repository contains the data, scripts and baseline co

51 Dec 17, 2022

Code, pre-trained models and saliency results for the paper "Boosting RGB-D Saliency Detection by Leveraging Unlabeled RGB Images".

Boosting RGB-D Saliency Detection by Leveraging Unlabeled RGB This repository is the official implementation of the paper. Our results comming soon in

8 May 22, 2022